Jan 1, 2005 - services to the University of California and delivers a dynamic ... Garner Valley, CA Acoustic and Seismic Experiments ... 9 sensor (6 tri-axial and 3 bi- axial), 8 of them on the perimeter of a 100 feet square, and the last.
Center for Embedded Network Sensing University of California
Title: Localization based on acoustic and seismic array processing Author: J.Z. Stafsudd; S. Asgari; C.E. Chen; A. Ali; R. E. Hudson; F. Lorenzelli; et al. Publication Date: 01-01-2005 Series: Posters Publication Info: Posters, Center for Embedded Network Sensing, UC Los Angeles Permalink: http://escholarship.org/uc/item/2wj9v34m Additional Info: BibTex @misc{ title={Localization based on acoustic and seismic array processing }, author={J.Z. Stafsudd, S. Asgari, C.E. Chen, A. Ali, R. E. Hudson, F. Lorenzelli, K. Yao, and E. Taciroglu}, abstract={In this poster, we consider the analysis, implementation, and application of wideband sources using both seismic and acoustic sensors. We use the AML algorithm to perform acoustic DOA. For non-uniform noise spectra, whitening filtering was applied to the received acoustic signals before the AML operation. For short-range seismic DOA applications, one method was based on eigen-decomposition of the covariance matrix and a second method was based on surface wave analysis. Experimental estimation of the DOAs and resulting localizations of a source generated by striking a heavy metal plateby a hammer on the ground using the acoustic and seismic signals individually and jointly are reported.}, url={http://research.cens.ucla.edu/pls/portal/ url/item/0425C02699AFBE86E0406180528D0AED}, year={2005}, } Abstract: In this poster, we consider the analysis, implementation, and application of wideband sources using both seismic and acoustic sensors. We use the AML algorithm to perform acoustic DOA. For non-uniform noise spectra, whitening filtering was applied to the received acoustic signals before the AML operation. For short-range seismic DOA applications, one method was based on eigen-decomposition of the covariance matrix and a second method was based on surface wave analysis. Experimental estimation of the DOAs and resulting localizations of a source generated by striking a heavy metal plateby a hammer on the ground using the acoustic and seismic signals individually and jointly are reported.
eScholarship provides open access, scholarly publishing services to the University of California and delivers a dynamic research platform to scholars worldwide.
Center for Embedded Networked Sensing
Localization Based on Acoustic and Seismic Array Processing J.Z. Stafsudd1, S. Asgari1, C.E. Chen1, A. Ali1, R.E. Hudson1, F. Lorenzelli1, K. Yao1, and E. Taciroglu2 Electrical Engineering Dept., UCLA1 , Civil and Environmental Engineering Dept., UCLA2 Introduction: Joint Acoustic and Seismic Source Localization
Acoustic DOA Estimation / Source Localization • •
Seismic DOA Estimation / Source Localization
A new version of Approximate Maximum Likelihood (AML) • algorithm is used to estimate the direction of arrival (DOA). • Differences between old version and new version of AML: 1-channel whitening : is used to reduce the effect of the non_uniformity of the noise. • 2-Better Freq. bin selection :Since the useful information are usually stored in the higher frequencies, we put more weight in the higher frequencies of the received data. (when we are computing the • criterion to choose a sub set of frequencies for further processing).
Seismic Event Detection Via Sample Covariance Matrix Utilized Previously Published DOA Estimation Scheme Developed for Long-Range Seismic Data Polarization analysis applied to short-range data
Novel Short-Range Techniques for Seismic DOA Estimation Surface Wave Analysis utilizes the unique characteristics of short-range seismic signal
Two Optimization Techniques for Seismic Source Localization from Estimated DOAs TLS criterion and L1 criterion
Problem Description: Locate a Source Generating both Acoustic and Seismic Signals with with Acoustic and Seismic Sensor Arrays
Garner Valley, CA Acoustic and Seismic Experiments
Acoustic and Seismic Sensor Array Equipments
•
Seismic Sensors Acoustic Arrays Metal Plate
18 feet 9
8
7
50
6
5
4
0
3
2
1
18 feet
100
24.6 feet
• • •
Location of Acoustic Arrays, Seismic Sensors, and Metal Plate 150
23 feet
•
Acoustic and Seismic Field Measurements
• Metal hammer struck a heavy metal plate at the marked location Acoustic and Seismic Sensor Map Acoustic Array Map
8 low cost Behringer XM200S microphones Each array consists of 4 microphones 1 meter apart in a square 2 acoustic arrays The output of 8 microphones are sent to the Presonus Firepod 8 channel 94bit/96kb firewire-based recording sys. The recording system synchronizes the acoustic signals and sends to a PC for AML-based DOA/loc. Algorithm Episensor tri-axial/bi-axial accelerometer sensors Accelerometers have wide frequency/amplitude ranges, and wide dynamic range Outputs of accelerometers are fed to the low power, high resolution Quanterra Q330s recording systems 9 sensor (6 tri-axial and 3 bi- axial), 8 of them on the perimeter of a 100 feet square, and the last one in the center
Y (feet)
• • • •
19.7 feet −50 −50
0
50 X (feet)
100
24.6 feet
150
Proposed Solution: Acoustic and Seismic Source Localization Performed Separately and and Fused for Final Source Localization Results Acoustic Event Detection Received hammering signal in time domain
Seismic Event Detection PSD of the noise
• Form sample covariance matrix from data of widow length N at each sensor • The size of the eigenvalues of the sample covariance matrix is a good indication of the presence of an event of interest • The simple eigen-decomposition procedure can be performed on sliding time windows through the data record to find signification events of interest • The data record with 6 significant hammer strikes was selected for analysis Eigen Value Progression As a Function of Window Number
Original Hammer Strike Signal, Sensor 1 Z Direction
−3
0.015
−4 0.01
−5
Log of Eigenvalues
Signal Strength
0.005
0
−6
−7
−8
−0.005
−9 −0.01
Eigenvalue #1 Eigenvalue #2 Eigenvalue #3
−10
−0.015
0
5
10
15
20 Time (sec)
25
30
35
−11
40
• The eigenvalue plot clearly indicates the 6 peaks above happened
10-4
0
50
100
150 200 Window Number
250
300
350
is where the 6 significant hammer strikes
Seismic DOA Estimation via Polarization Analysis
Received hammering signal in Frequency domain
Normalized signal in Frequency domain
• The largest eigenvalue of the covariance matrix corresponds to the average energy of the strongest seismic mode polarized in the direction of the corresponding eigenvector • Same can be said for the second and the smallest eigenvalues and their corresponding eigenvectors • Rayleigh wave is elliptically polarized, occupying two orthogonal directions while Love wave is rectilinearly polarized in the direction orthogonal to the direction of propagation • The estimated DOAs have a 180-degree ambiguity
Seismic DOA Estimation via Surface Wave Analysis • All motions in the vertical direction are due to Rayleigh wave • The other component of Rayleigh wave and the direction of motion of the Love wave are perpendicular to each other, and are both in the ground plane • Rayleigh wave exhibits an elliptical motion in the plane made up by the vertical direction (z) and the direction of propagation (p) • The velocity in the vertical direction (z) and the acceleration in the direction of propagation (p) are in phase and proportional to each other • The projection of the integral of the vertical (z) acceleration onto the ground plane defines the direction of propagation • The estimated DOAs have no ambiguity
Seismic Source Localization with DOA Estimates • L1 formulation: Minimize the sum of distances between an arbitary point to the estimated DOA line • TLS formulation: Minimize the sum of squares of distances from an arbitary point to the estimated DOA line • Weighted with the product of uncertainty in the ith estimated DOA and the distance between the airbitary point and the sensor i Source Localization Results from Surface Wave Analysis, Weighted TLS
Source Localization Results from Surface Wave Analysis, Weighted L1
Source Localization Results from Covariance Analysis, Weighted L1 150
150
150
Source Localization Results from Covariance Analysis, Weighted TLS 150
1
2
4
5
7
8
100
3
1
2
3
4
5
6
7
8
9
1
100
2
1
2
3
4
5
6
7
8
100
3
50
9
0
50 X (feet)
6
8
9
Sensor Positions True Source Location Directions At Each Sensor Estimated Source Location
100
150
−50 −50
50
0
50 X (feet)
100
150
Sensor Positions True Source Location Directions At Each Sensor Estimated Source Location
−50 −50
−50 −50
0
50 X (feet)
100
0
50 X (feet)
150
Acoustic and Seismic Source Localization Results, Hammer Strike Case A 180 Seis. Snsr Loc. Acou. Snsr Loc. Metal Plate Loc. Seis. SrcLoc Acou. SrcLoc Fusion SrcLoc
160
Fusion results are obtained by averaging the x- and y-coordinates of both acoustic and seismic source localization results
140
120
X coordinate
True Source Location
Acoustic Source Localization Result 75
Seismic Source Localization Result 78.4
Fusion Source Localization Results 80.6
2
1
3
100
Y (feet)
Source Localization Results in Feet
79.5
80
1
60
4
5
40
6
2
20
Y coordinate
25
23.8
26.7
25.2
7
8
9
0
−20 −20
0
20
40
9
0
Sensor Positions True Source Location Directions At Each Sensor Estimated Source Location
Fusion of Acoustic and Seismic Source Localization Results •
5
7 0
Sensor Positions True Source Location Directions At Each Sensor Estimated Source Location
−50 −50
4 50
0
0
Y (feet)
50
Y (feet)
Y (feet)
Y (feet)
100
6
60 X (feet)
80
100
UCLA – UCR – Caltech – USC – CSU – JPL – UC Merced
120
140
100
150
